RGBD-DSO Direct Sparse Odometry with RGB-D Cameras for Indoor Scenes

1. Related Papers

• RGB-D DSO: Direct Sparse Odometry with RGB-D Cameras for Indoor Scenes, Yuan Z, Cheng K, Tang J, Yang X, In IEEE Transactions on Multimedia, 2021

2. Installation

    git clone https://github.com/HustCK/RGBD-DSO.git

2.1 Required Dependencies

2.1.1 Suitesparse

Install with

    sudo apt-get install libsuitesparse-dev libboost-all-dev

2.1.2 Eigen3

Eigen 3.2.8, Follow Eigen Installation.

2.1.3 OpenCV

OpenCV 2.4.9, Follow OpenCV Installation.

2.1.4 Pangolin

Pangolin, Follow Pangolin Installation.

2.1.5 ziplib

Install with

    sudo apt-get install zlib1g-dev
    cd dso/thirdparty
    tar -zxvf libzip-1.1.1.tar.gz
    cd libzip-1.1.1/
    ./configure
    make
    sudo make install
    sudo cp lib/zipconf.h /usr/local/include/zipconf.h

2.2 Build

    cd RGBD-DSO
    mkdir build
    cd build
    cmake ..
    make -j4

3. Usage

3.1 Dataset Format

Let's take TUM RGB-D as an example.

    <sequence folder name>
        |____________rgb
        |____________depth
        |____________associate.txt

If you are using other datasets, pleasr adjust the file directory and format as described above.

3.2 Run

If you use the same datasets as in this article, run it directly with the following instructions:

    bin/dso_dataset \
        files=<sequence folder name> \
        calib=<RGB-D DSO path>/calib/<dataset name>/calib.txt \
        preset=0 \
        mode=1

For more details on configuration parameters, see Direct Sparse Odometry.

4. Acknowledgement

This work is implemented based on Direct Sparse Odometry. Thanks to J. Engel et al., who open source such excellent code for community.